The Unlikely Middle: Overestimation of Most and Least Likely Outcomes

Assessing of the probability of future outcomes is central to most forward-looking behaviors, from an investor assessing the risk of a portfolio to a consumer assessing how a currently available price is likely to compare to future prices. There is a large literature on how people make probability judgments and risk estimates and the degree to which such estimates are biased (Tversky and Kahneman 1973, Gigerenzer 1994, Slovic 2000, Loewenstein et al 2001, Krizan and Windschitl 2007, etc.). One limitation in this literature has been that probabilities are often estimated in isolation or merely rated based on subjective risk. In this paper, we present a new task in which participants jointly estimate the probability of mutually exclusive events that can be objectively calculated, but which is difficult to do so. The goal is to test the accuracy of people’s probability judgments and whether biases in perceived probability would extend to this context. In the studies, a simple and intuitive situation is described to participants, e.g. “In the game, you would flip a fair coin 10 times and you would show the experimenter the result each time. Each time that it comes up heads, you win $1, and each time it comes up tails, you get nothing.” Participants are then asked to estimate (using different elicitation methods) the probability of each of the possible outcomes (getting $0 to $10) simultaneously. By jointly eliciting the probabilities, and enforcing the provision that they must add up to 100%, we avoid the separate issue of subadditivity in probability judgments (Tversky and Koehler 1994).